Audio signals may include both desired components, such as a user's voice, and undesired components, such as noise. Noise removal (or cancellation) attempts to remove the undesired components from the audio signals. One implementation of noise removal is dual microphone noise cancellation, where a first microphone is used to pick up primarily a desired signal (e.g., the user's voice) and a second microphone is used to pick up primarily an undesired signal (e.g., a noise signal, such as background noise). The dual microphone cancellation system may remove noise by subtracting the audio signal picked up by the second microphone from the audio signal picked up by the first microphone. This and other noise cancellation techniques have various drawbacks. For example, this noise cancellation technique does not perform well if the geometry of the audio source versus the noise source is not fixed or known. These and other drawbacks are addressed in this disclosure.